Location Tracker

Environment

  • OS: Ubuntu 16.04
  • Language: Python3.5

Install

For gpu version:

  1. Install cuda and cudnn. Their version should match tensorflow-gpu with version 1.8.0
  2. Install python package: pip3 install -r requires.txt

For cpu version:

  1. Install python package: pip3 install -r requires.txt
  2. Remove tensorflow-gpu and install tensorflow: pip3 uninstall tensorflow-gpu & pip3 install tensorflow==1.8.0

Run program

  1. Generate all temporary file: ./build.sh
  • It will generate a directory named tmp which contains all temporary files.
  1. Train model: ./train.py my_model
  • It will spend about 25 minutes to train a model with 5-fold cross validation in a nvidia 1080ti GPU.
  1. Generate result: ./eval.py my_model
  • It generate a directory named result. And there are two files in the directory.
    • result.txt: output file of this homework
    • check.txt: the file that show the sorted classification probability of each user